Geographical Variation of Incidence of Chronic Obstructive Pulmonary Disease in Manitoba, Canada

• Main Authors: Mahmoud Torabi, Katie Galloway
• Format: Article
• Language: English
• Published: MDPI AG 2014-07-01
• Series: ISPRS International Journal of Geo-Information
• Subjects: bayesian computation; chronic obstructive pulmonary disease; geographic epidemiology; prediction; random effects; spatial cluster detection
• Online Access: http://www.mdpi.com/2220-9964/3/3/1039

We aimed to study the geographic variation in the incidence of COPD. We used health survey data (weighted to the population level) to identify 56,944 cases of COPD in Manitoba, Canada from 2001 to 2010. We used five cluster detection procedures, circular spatial scan statistic (CSS), flexible spatial scan statistic (FSS), Bayesian disease mapping (BYM), maximum likelihood estimation (MLE), and local indicator of spatial association (LISA). Our results showed that there are some regions in southern Manitoba that are potential clusters of COPD cases. The FSS method identified more regions than the CSS and LISA methods and the BYM and MLE methods identified similar regions as potential clusters. Most of the regions identified by the MLE and BYM methods were also identified by the FSS method and most of the regions identified by the CSS method were also identified by most of the other methods. The CSS, FSS and LISA methods identify potential clusters but are not able to control for confounders at the same time. However, the BYM and MLE methods can simultaneously identify potential clusters and control for possible confounders. Overall, we recommend using the BYM and MLE methods for cluster detection in areas with similar population and structure of regions as those in Manitoba.